Feature Selection Method Based on High-Resolution Remote Sensing Images and the Effect of Sensitive Features on Classification Accuracy

Abstract

With the advent of high spatial resolution remote sensing imagery, numerous image features can be utilized. Applying a reasonable feature selection approach is critical to effectively reduce feature redundancy and improve the efficiency and accuracy of classification. This paper proposes a novel feature selection approach, in which ReliefF, genetic algorithm, and support vector machine (RFGASVM) are integrated to extract buildings. We adopt the ReliefF algorithm to preliminary filter high-dimensional features in the feature database. After eliminating the sorted features, the feature subset and the C and γ parameters of support vector machine (SVM) are encoded into the chromosome of the genetic algorithm. A fitness function is constructed considering the sample identification accuracy, the number of selected features, and the feature cost. The proposed method was applied to high-resolution images obtained from different sensors, GF-2, BJ-2, and unmanned aerial vehicles (UAV). The confusion matrix, precision, recall and F1-score were applied to assess the accuracy. The results showed that the proposed method achieved feature reduction, and the overall accuracy (OA) was more than 85%, with Kappa coefficient values of 0.80, 0.83 and 0.85, respectively. The precision of each image was more than 85%. The time efficiency of the proposed method was two-fold greater than SVM with all the features. The RFGASVM method has the advantages of large feature reduction and high extraction performance and can be applied in feature selection.

Separating Built-Up Areas from Bare Land in Mediterranean Cities Using Sentinel-2A Imagery

“…An increased number of features, such as the image bands, generally improves the accuracy; however, it increases the amount of training samples required [24]. Considering the limited training sample availability in most conditions and the nonlinear response of the LCU classes across several bands, which is known as the "Hughes effect" [25], there is a need for a method to select a subset of relevant features from the original dataset to improve the classification process and achieve a dimension reduction [26].One alternative to overcome the above-mentioned problem is to perform linear transformation methods (such as principal components analysis (PCA) and independent component analysis (ICA)), or nonlinear algorithms (such as locality adaptive discriminant analysis (LADA) and multiple marginal fisher analysis (MMFA)), to remove the correlations and higher-order dependences in the image bands and use the produced components as input data for classification, to simplify and improve the process. The linear methods have been widely applied on multispectral data, however, nonlinear methods are generally applied on hyperspectral test data or natural image-based applications, such as face recognition [27][28][29][30].Remote Sens.…”

“…An increased number of features, such as the image bands, generally improves the accuracy; however, it increases the amount of training samples required [24]. Considering the limited training sample availability in most conditions and the nonlinear response of the LCU classes across several bands, which is known as the "Hughes effect" [25], there is a need for a method to select a subset of relevant features from the original dataset to improve the classification process and achieve a dimension reduction [26].…”

Decoding and Systematization of Medical Imaging Features of Multiple Human Malignancies

“…The feature extraction process, especially image filtering and texture matrix calculation, was markedly affected by the different methods used. These flaws may incur serious bias, leading to unrepeatable studies and invalid conclusions (14)(15)(16)(35)(36)(37). Recently, concerns and possible solutions have been proposed to this challenge (13,23,38).…”

A novel remote sensing detection method for buildings damaged by earthquake based on multiscale adaptive multiple feature fusion

“…High-resolution images provide clearer boundary contours, more striking textural information, richer colour hues and more explicit spatial information then the medium-or low-resolution counterparts (Cleve et al 2008). The extraction of building damage information from feature-assisted high-resolution remote sensing images is divided into three categories, namely, pixel-based feature recognition (Blaschke et al 2014;, object-oriented recognition (Zhou et al 2018) and image change detection before and after the disaster (Huang et al 2018). In the absence of pre-disaster remote sensing images, researchers automatically extract collapsed house information from single-phase images after the disaster and extract damage information through models or classifiers (Shafique et al 2011;Xie et al 2016;Adriano et al 2019;Li et al 2019).…”